Harold Reynolds displays his genius

[TheRealMattKain]

One more time. You said:

I like players who take their walks but that does not mean there aren't situations where I would rather the player made an out attempting to hit the ball than take a walk and hope an inferior player can score the run.

I still say YIKES. I know what you're trying to say here, but in an individual case, I will never take an out over a walk. That's just silly. I think what you really mean is that given certain situations where the outcome is unknown, you would sometimes rather have the player try to get a hit than take a walk.

Not sure how you think one person will win you anything since your hypothetical team could be made up of Bill Hall types but I guess you have an answer for how that won't hurt you if you have someone who is on first when he strikes out.

Now this is just getting frustrating. Where did this whole "Bill Hall types" thing come from? Nowhere in any of my posts is anything like this mentioned. You're totally losing sight of the argument. Let me try one more time to rephrase and refocus the debate.

1. Poster claimed that a hitter who walks every time would be far more valuable than anyone else in baseball.

2. You countered by saying this is not true. Your argument being that several current players would have a higher OPS than a theoretical player that walks every single plate appearance. You said: "Taking a walk every at bat would have been good enough for fourth in OPS last season. I think Chipper Jones, Albert Pujols and Manny Ramirez were more valuable last year than the player you described."

3. I countered this by claiming that OPS is not a good tool to use to analyze this theoretical player. OPS overvalues slugging and undervalues on base percentage. Since we're talking about a player that walks every time, this is, in fact, the ultimate extreme case of where OPS does not successfully measure a player's offensive value. I claim that the true way to measure this player's value is to try plugging him into a lineup simulator and see how much additional offense he will produce. The results speak for themselves (I assume you have not tried a lineup simulator, correct?). Using a lineup simulator, we see that a player that walks every time will result in the team scoring more runs than a Pujols, Manny, etc.

Going back to your trading me one guy who will be get to first base, and only first base, 5 times a game for one of those three I think I'll take you up on that. My guy would be able to score on his own your needs help every time. If both our teams are full of Bill Halls but I have Pujols and you have Jonny take your base who do you think will win more games?

Me.

To quote BrewJihad:

"1.000 OBP leading off, .200/.350 hitters in the rest of the lineup: 3.736 runs per game
Pujols hitting 3rd, .200/.350 hitters in the rest of the lineup: 2.530 runs per game"

There you go. My team would average more runs per game than yours. It's not even close.

Getting back to your side (assuming that you still disagree), I must assume one of the following:

1. You are simply failing to follow the debate that you initiated (I admit that there are many different thoughts and arguments going on simultaneously in this thread so it could be a bit confusing).

2. You think computer programs that simulate run scoring using different lineups are bunk (if so, please explain).

3. You believe computer programs that simulate run scoring have some merit, but you still believe that in this case OPS is a better indicator of run scoring (if so, please explain).

4. You have not tried and/or possibly do not understand lineup simulators. If so, nothing wrong with that. Try one of these links to learn more:
http://www.sportsquant.com/baseballapp.htm
http://www.baseballmusings.com/cgi-bin/LineupAnalysis.py

[Thurston Fluff]

I think what you really mean is that given certain situations where the outcome is unknown, you would sometimes rather have the player try to get a hit than take a walk.

 

I don't think there is ever a circumstance where the outcome is known so I thought that was a given. That is why I said attempting to get a hit. If you are attempting something the outcome is unknown by definition. I think our problem here is semantics.

 

2. You think computer programs that simulate run scoring using different lineups are bunk (if so, please explain).

 

I don't think they are bunk but i do think they are only as good as the formulas they use to predict outcomes. The formulas used are based off real life outcomes not some hypothetical unrealistic what if situation.

I do not know how well a formula that was based off actual outcomes works when you use them to run a simulation of an extreme, unrealistic, circumstance. I have a feeling the basis for the calculations would change if those extreme circumstances really happened and were included. There certainly should be room for debate on their accuracy to assess something they were not meant to assess. Not sure if I'm explaining my misgivings very well but simply put they work well enough in normal usage but not sure if it follows that it works the same in unrealistic ones.

[TheRealMattKain]

I don't think they are bunk but i do think they are only as good as the formulas they use to predict outcomes. The formulas used are based off real life outcomes not some hypothetical unrealistic what if situation.

I do not know how well a formula that was based off actual outcomes works when you use them to run a simulation of an extreme, unrealistic, circumstance. I have a feeling the basis for the calculations would change if those extreme circumstances really happened and were included. There certainly should be room for debate on their accuracy to assess something they were not meant to assess. Not sure if I'm explaining my misgivings very well but simply put they work well enough in normal usage but not sure if it follows that it works the same in unrealistic ones.

I think I follow you. This is simply where we disagree. Your fear is that programming limitations or potential bugs that may manifest in an extreme case cause you to mistrust the run scoring/lineup simulators. Maybe as a software engineer I tend to have a little more faith in these run scoring/lineup simulators. Sure, if plugging in my fictional player generated some sort of ridiculous run-scoring outcome (negative runs, 0 runs, infinite runs, etc.) then I would probably doubt the results as well. However, in my opinion the run scoring averages generated by these programs seem pretty much right in line with what I would assume a player that walks every at bat would add. What is interesting here, however, is that my theoretical player has no unknowns. If anything I would expect the simulator to work even better in this case. The outcome for my player is always known. My player gets on base every single time and my player never makes an out. That's huge. Think about it for a second. Given 650 plate appearances a year, a superstar with a .420 on base percentage gets on base 273 times and makes 377 outs. My player gets on base 650 times and makes 0 outs. That, my friend, is a massive difference (even if your guy hits doubles, homers, etc.).

 

Sure, we can come up with all sorts of anecdotal scenarios (what if each time my player gets a walk, the next three batters strike out?). I can do the same. What about when Pujols slumps and goes 0-20? What about when Pujols strikes out with the bases loaded to end an inning? This is exactly the point of the simulator. It can run through thousands of scenarios to create an accurate gauge of what we would expect, on average, from a given lineup. Some of those scenarios might be the ones you mentioned (where my player gets a walk and it does not result in a run because the following hitters failed to bring him home). That's the advantage of letting a computer run a simulation (I feel silly even typing this). We can feed in a series of probabilities and let it compute a likely outcome. Repeat this thousands of times, and we start to get really, really close to what we would expect to happen in real life. .

 

Your general line of thinking seems to be anti-statistic. Why do you trust OPS at all? Given a player with a 1.000 OPS how do we know that he won't strike out 100 times in a row? Or he could hit into 30 double-plays in a row, right? Doesn't that mean we should mistrust OPS as well?

[MJLiverock]

Between a race car driver and a person with a masters in structural engineering, I'll assume the engineer is right

Best comment of the day, always defer to the judgement of engineers when dealing with anything...

This one made me chuckle a little because I just got back from a two day course on oil and gas exploration taught by an engineer and he said never trust and engineer if the outcome involves risk, (i.e. uncertain outcomes) because they are too risk averse to make the correct decision.

[Thurston Fluff]

Your general line of thinking seems to be anti-statistic.

 

I'm not anti-statistic at all. I just question how well something can work when it is calculating something that was not incorporated in the calculation. Walks occur for a variety of reasons. Some of them because the pitcher is tired or just not very good. Some walk because they are setting up a double play. Some walk because they are so good the pitcher would rather take his chances with the next guy. Some walk because the next hitter is so bad the pitcher decides the chances of scoring are less with the inferior hitter even if there is a man on first. The simulators take all the reasons for a walk and lump them into one calculation and come out saying a walk is worth X number of runs regardless of situation. That is perfectly fine when calculating normal occurrences. Most hitters will end up walking for most if not all of those reasons sometime or another.

But here is the sticking point. There is no part of the equation that incorporates what happens when a player cannot not walk. It is hard to know how that would turn out. Does the pitcher not pitch to him and thus is less likely to show the next batter his stuff? Does the pitcher hit the batter in the leg so he can't run as well? Does the pitcher save himself for the next hitter and not expend any energy on that guy? Does the pitcher throw pick off throw after pick off throw until he finally gets him? Does the opposing team game plan how to approach the next batter in the lineup differently with the knowledge there will always be a man on first when he comes up?

Who knows. Since it never happened and the calculations are based off real happenings the simulators don't have that information available for their calculations.

 

Why do you trust OPS at all? Given a player with a 1.000 OPS how do we know that he won't strike out 100 times in a row? Or he could hit into 30 double-plays in a row, right? Doesn't that mean we should mistrust OPS as well?

 

However unlikely it is at least possible for a player to do all the things you said. They are all based on how well, or poorly, the individual player performs. over the long haul player streaks evens out. Nor does any of those things change the criteria of the calculations. The concept of a player walking every time he comes up is not. It is based on both his ability and the pitchers willingness to throw him non strikes. Any pitcher not named Jorge De la Rosa would throw strike after strike until the player proved he could hit it.

 

Really this is all academic and probably pretty boring for anyone stuck reading it so, as fun as it has been, maybe it's time to move one. But thank you for the debate it was compelling.

 

All in all I think Reynolds was pretty far off base by any measure. That doesn't mean there are not times a hitter should try to get a hit vs take a walk. Just not all that many.

[TooLiveBrew]

"A nice play by a very good centerfielder, McLouth -- a GG award winner last year -- runs down the shot off of A-Rod's bat"

 

That's a quote from NYY pbp announcer Michael Kay. Nate McLouth just made an average play in CF, and didn't even get a great jump (nor was the ball hit particularly well). This is the problem I have to deferring to most media types & former ballplayers. They often times are just wrong. 'Because I said so!' is just a bad reason to believe anyone.

[TheRealMattKain]

I just question how well something can work when it is calculating something that was not incorporated in the calculation. Walks occur for a variety of reasons. Some of them because the pitcher is tired or just not very good. Some walk because they are setting up a double play. Some walk because they are so good the pitcher would rather take his chances with the next guy. Some walk because the next hitter is so bad the pitcher decides the chances of scoring are less with the inferior hitter even if there is a man on first. The simulators take all the reasons for a walk and lump them into one calculation and come out saying a walk is worth X number of runs regardless of situation.

 

 

You do not understand how a run-scoring simulator works. It does not "lump them into one calculation" or say "a walk is worth X number of runs regardless of situation." A probability is assigned to the likelihood of an event occurring based on prior outcomes. In our case these prior outcomes are things like batting average, slugging percentage, on base percentage, etc. For Willie Walks-Every-Time there is a 100% chance that a walk will occur because Willie walks every single plate appearance. Your claim that it is "not incorporated in the calculation" is completely incorrect. The complete opposite is now true. We are now incorporating an absolute known outcome rather than a probability into the simulator (again, there is a 100% chance of a walk occurring). If the outcome for all 9 lineup spots was known, then the simulator would be 100%, picture-perfect accurate. Do you see what I'm saying? By adding an event with a known outcome, the simulator actually gets closer to being correct (not more incorrect as you assume). I shouldn't have to give a stats/combinatorics lecture here - this is nothing more than basic probability.

 

There is no part of the equation that incorporates what happens when a player cannot not walk. It is hard to know how that would turn out. Does the pitcher not pitch to him and thus is less likely to show the next batter his stuff? Does the pitcher hit the batter in the leg so he can't run as well? Does the pitcher save himself for the next hitter and not expend any energy on that guy? Does the pitcher throw pick off throw after pick off throw until he finally gets him?

 

 

Now you're just being absurd. The effect of these potential side-effects is so minor that it is irrelevant compared to the huge gains that a walk every plate appearance would create. Again, I see that this really boils down to your not understanding how a lineup/run scoring simulator works. A set of event probabilities based on player statistics is used to determine a likely outcome. None of your above arguments would alter the set of event probabilities for the remaining 8 players in the lineup enough to eliminate the tremendous advantage having a player walk every single time would produce.

The bottom line is this: if you fail to see how massively valuable a theoretical player who walks every plate appearance is, then you also fail to see the importance of on base percentage. You can quibble and come up with as many silly "what ifs" as you want. If you really don't see how beneficial a perfect 1.000 on base percentage (even if it's all generated via walks) is, then you're really missing the importance of on base percentage in general. As has been said several times in this thread already: not making outs in baseball is huge.

Anyway, I see that our understanding of this is radically different. I suggest playing with the simulators a little bit and seeing what kind of results you get (or maybe reading up on probability theory since it has a pretty useful application for baseball - my fantasy team can attest to that). Thanks for a lively discussion.

EDIT: fixed some minor typos

[TheCrew07]

There's two different things really being discussed here, and I've made points on both sides in the past.

 

First, Sabermetrics have not changed the way baseball is played, they have changed the way we evaluate players and some strategy in the game, but the nuts and bolts of baseball remains the same as it always has been. Pitching is about location, hitting is about selectivity. Over 100 years of trial and error people have figured out what it takes to get other players out and how to hit for high average. The reason many players don't make good announcers is that many times they know just know something to be true, but don't know how to explain it. Call it the "zone" or whatever, they just somehow instinctively know what to do and have the talent to execute it. Nothing in sabermetrics is going to teach a player how to hit a ball harder/more consistently or control their pitches/location better.

 

It's pretty obvious to me why these baseball players get their dander up when the "nerds at bf.net" (as we've been affectionately referred to in the past) start telling them what's right and wrong in baseball, statstical analysis really has nothing to do with what happens in a particular game. For example my least favorite overused stat is BABIP... it's treated like the "cause" in a cause and effect scenario, but it's truly an "effect". Over a smaller sample size it may indicate some luck, but no one is lucky or unlucky for an entire year, at some point the players deserve credit or blame... A hitter who has a BABIP over .360 for a season isn't lucky, he's talented and making great contact, thus the resulting high BABIP. Just like a pitcher who's BABIP is high is likely leaving pitches up in the zone and catching too much of plate... he should get smacked around because if MLB hitters aren't connecting on those pitches, they don't belong in MLB in the first place. The whole notion of luck to me is horrible analysis... "I don't know how to account for all of the variables so it must be luck"...

 

The second point here is that players like to judge each other on some mix of talent and results. I'm fortunate enough to know people who've played both football and baseball professionally and nothing irritates them more than players with "million dollar tools and a 10 cent brain". It takes an abnormal amount of self confidence to make it as a professional athlete, often mistaken for arrogance, and in some cases they truly are arrogant. They don't have much use for analysis beyond what their eyes tell them, because when they were playing the game, they only had time to react to what they saw, there's not time for thoughtful reflection during a game, it just gets in the way. They were successful because 100s of little things that sabermetrics isn't able to quantify and statistitians dismiss as "noise" like gut feelings, hunches, and so on. Some players know what's going to happen before it does... how do you explain or account for that? I see their point, I think the sabermetric tools are much more meaningful to GMs than to the players, but I wish these players would take more time to understand and appreciate some of the very interesting work that's being done.

 

While I believe many current and former players, and in this case Renyolds, are too dismissive of advanced statistical analysis, I think the statistical community is just as arrogant and dismissive in regards to the truisms of baseball. There's a general trend towards dismissing general baseball wisdom as "old school cliche"... and while some of it is, like the value of small ball, much of it isn't... For example people around here like to make of fun of Yost for his grit and determination speech... but that's exactly what it takes to succeed in a sport where you fail more than you succeed, he wasn't wrong. If you don't have the proper attitude you won't succeed regardless of your talent, and it's mind boggling to me how dismissive statisticians tend to be in regard to the mental aspect of sports.

 

Taking it further to a pitcher location is everything, analysis of pitching should start and end with pitch location relative to the target, though I do like much of the pitch FX work that's being done... to a hitter seeing his pitch when he's looking for it is everything, if he gets the pitch he expects in a good location chances are he's going to hit it hard some place. It's easy to look at player who doesn't succeed and make a stastical arguement that he's just not very talented, like the case of Hall, when that couldn't be further from the truth. He's arguably the best athlete on the team who gets the least out of his ability... just because you aren't able to measure it statistically doesn't mean the little nuances in the game aren't the biggest factors in the outcome of a particular game, AB, or a player's career.

 

The recent Bill James article linked on BCB last week was fascinating to me, I've always enjoyed reading his opinions on issues and he took shots at pitch counts and indirectly at PAP. 100 pitches is arbitrary... it's not based on anything scientific other than being a round number.. Nolan Ryan is trying to take on what's recently become conventional wisdom and push the pitchers in his organization out past 100 pitches. How do we know the true threshold isn't 110 pitches, or 130? What about as a pitcher matures, shouldn't his pitch counts go up? Is what's best for a developing arm the same as what's best for a mature arm? Shouldn't we be trying to maximize the value of the best pitchers on the team instead replacing them with inferior pitching talent from the bullpen unnecessarily? One of the things that stuck with me from the article was a quote from a GM about the culture of baseball... that we've created a culture in baseball where a pitcher thinks if he goes 6 scoreless in 100 pitches his night should be over, when that's not necessarily the case.

 

I find the whole counter culture movement fascintating as people begin to realize that maybe their analysis has gone too far. In Brewer terms I think a 100 pitch limit is a very good idea for Yo this season given his recent history, but if Suppan could be effective out to 120-130 pitches, I wouldn't have a problem with him throwing that much.

[Thurston Fluff]

It does not "lump them into one calculation" or say "a walk is worth X number of runs regardless of situation."

 

It does not differentiate between how a walk occurred. A walk given to the #8 hitter with a pitcher coming up and a man on third with two outs has a better chance of preventing a run from scoring than not walking the guy. Yet it is assigned the same value as a walk issues to the lead off man in the first inning. For the most part that is fine since over a regular season most of those circumstances sort of even themselves out. What is not fine is to add about 500 walks from a source completely out of the blue that has never been part of the analysis and just expect it to accurate.

Maybe example will be useful outside of baseball stats will be helpful to my point. I'm sure you have seen those alcohol analyzers. You add your weight and drinks you've had and the period of time you drank them in and it tells you your blood alcohol level. For a normal person it works fine. For an alcoholic it does not. That is because hardcore alcoholics get drunk at a different rate than normal.

The same thing is happening with the fake player who walks all the time. It is an outlier that may not compute normally. I don't know if it would mess with it but I don't think you should be as confident as you are it will work the same. I don't think any of them were accepted as accurate until they actually proved themselves to be so. In the circumstance where an unusual situation arises I think it has to be shown to work as accurately as it does in normal situations.

 

 

Now you're just being absurd. The effect of these potential side-effects is so minor that it is irrelevant compared to the huge gains that a walk every plate appearance would create. Again, I see that this really boils down to your not understanding how a lineup/run scoring simulator works.

 

Maybe stop with the absurd stuff it does nothing to help make your point. I think I do understand. I have used them. They make a lot of sense if you want to see how a player like 3TO would produce over a full season or how a player would do if he maintained his current rate of production. What I think you are missing is they are only as good as the statistical analysis they are basing their outcome on. That means they should be used in circumstances at least roughly similar to the data used to make their analysis. I don't think it's absurd at all to think predictive tools based on past results work best when used in similar circumstances to the data gathered in making those analysis. Nor is it absurd to question the validity of using them in predicting what would happen in outlier cases. Heck there are cases each year where a player or two do much better/worse than these same formulas tell us he would so they have their hands full doing normal stuff 100% let alone weird situations.

 

Thecrew07 did you happen to see the piece on Bill James on 60 minutes? I thought it was well done. One of the questions was about clutch hitting. Judging from how he answered that question it appears he is revisiting clutch hitting again. It would suck to finally accept there is no such thing only to have him come out and prove there is.

[TooLiveBrew]

I think the best way 'clutch' is explained is players being able to perform at their normal level in higher-leverage situations. I don't think it's worth worrying about much in either direction (good/bad), since most professional athletes have long since separated themselves from those that fold under pressure.

[Thurston Fluff]

Probably right TLB but hearing James say he's revisiting that made me wonder. It has to be one of the hardest things to quatify since clutch is a subject ive term to begin with. I'd be interesting to see his new take on it if there is one.

[MJLiverock]

The hard part for hard core statisticians to accept clutch results is going to be arriving at the 1000's of points of data for each player in a "clutch" situation. Results that show clutchiness may be dismissed as small samples, espcially for individual players.

[knowitall]

Walks are just a stat - but they do show that a player has patience at the plate and is probably getting his pitch to hit. Low walk totals usually show impatience at the plate or a mechanical flaw by the hitter such as head or body movement, which usually means not seeing the ball as well and swinging at pitcher's pitches. Some players should take more walks because they are not great hitters and need to get on base. Other players need to be more aggressive because they are run producers with higher averages and slugging. As Harold Reynolds states, the situation should dictate a players approach.

[TheCrew07]

Thecrew07 did you happen to see the piece on Bill James on 60 minutes? I thought it was well done. One of the questions was about clutch hitting. Judging from how he answered that question it appears he is revisiting clutch hitting again. It would suck to finally accept there is no such thing only to have him come out and prove there is.

Sorry for the slow response, I intentionally stayed away from this thread for a couple of days.

 

I missed the piece, I don't watch much regular television anymore except for sports. Truth be told if I'm not watching sports or a movie, I'm probably watching the history or discovery channel. I'm not sure what exactly it is inside me but I've always been drawn towards military programming so I'm typically watching that type of program or playing a game similar in nature.

[Thurston Fluff]

No that I want to bring this whole thing up again but I did find a perfect example of a time swinging for a hit instead of taking working a walk would have been called for. Last night JJ Hardy was up in the late innings, 8th I beleive. He was already 4 for 4 and they just showed a graphic of each ball he hit. Each one came on a different pitch in a different location. One of which was well out of the strike zone. IOW he was seeing the ball well. There was a man on second with 2 outs and he had a 3-0 count. He took a get me over fastball over the middle of the plate for a strike. The next pitch was ball 4. The next batter made an out and 2 men were left on base.

I think a 3-0 count for a guy obviously seeing the ball well is one where the batter should be swinging if the ball is a hitters pitch like that one was. A walk in that instance was probably decreasing the odds of scoring more runs in the game. It was a reliever so there was no need to raise the pitch count he was going to be gone sooner than later anyway. As the game played out it would have been nice to have scored the extra run. Maybe then we wouldn't have had to use an extra reliever ourselves.

 

There are instances where the hitter should hit vs taking a walk. That was one of them.

[logan82]

I think many agree with you. I believe that I posted that the hitter should wait for a good pitch to hit. Obviously he got a good pitch to hit. I disagreed with "expanding the zone" except for maybe Braun who hits almost any pitch well.

[homer]

The hard part for hard core statisticians to accept clutch results is going to be arriving at the 1000's of points of data for each player in a "clutch" situation. Results that show clutchiness may be dismissed as small samples, espcially for individual players.

True, but the pro clutch crowd will seize on a handful of ABs as proof of clutch when it is anything but proven.

[mletto157]

Backupcatchers - Almost no one swings on 3-0 because even the best hitters will make an out 2/3 of the time; and the chances he can get on base via the walk were extremely high. I think that to swing or take the walk would depend largely on who follows them in the lineup though. In this case, Ryan Braun was hitting behind JJ and Prince after that. 100% of the time, I want as many people on base for Braun/Fielder as possible - so I think taking that walk was the right move. Now, if JJ was hitting 7th or 8th in the lineup I would agree with you because his chances of a hit far outweigh the chances of Kendall or the pitcher - especially in that count.